On Wed, May 27, 2015 at 9:53 AM, Chris J Arges chris.j.arges@canonical.com wrote:
On Wed, May 27, 2015 at 03:24:17PM +0200, Takashi Iwai wrote:
[Added more people to Cc]
At Wed, 27 May 2015 08:55:21 -0400, Adam Goode wrote:
The settings were not reset across a mem sleep (echo mem > /sys/power/state).
Hard power off / power on does make the settings get out of sync.
Ok did some testing on my Scarlett 18i8: suspend/resume -> settings not reset poweroff/poweron -> settings not reset unplug/replug usb -> settings not reset powercycle unit -> settings not reset
Is there a more consistent method to trigger this issue? In addition are you manually triggering 'alsactl restore/save' in this process?
I'll do some additional testing to see if I can trigger this.
Then maybe the problem is that the device keeps the old setting while the driver resets to the default. I vaguely remember that scarlett could save the default state in a persistent area, and the original driver patch had a kctl to trigger it. We didn't take it because it might be dangerous.
The original issue was this: Hey, as Tobias mentioned this is a HW saving (to the mixer's NVRAM) function used for using the mixer disconnected from a computer. We wouldn't want to continually write the NVRAM on every control update as I'm unsure of how many write-cycles the device is capable of.
So if we wrote to NVRAM on each control update we might wear out the memory very quickly. I think this functionality is more for using the device disconnected from a computer.
Takashi
Adam
On Wed, May 27, 2015 at 2:43 AM, Takashi Iwai tiwai@suse.de wrote:
At Tue, 26 May 2015 21:39:22 -0400, Adam Goode wrote:
Linux 4.0.2-300.fc22.x86_64
Hi,
I have the Scarlett 18i8 USB device that is supported by the new scarlett_mixer.c code. I am happy to say that the mixer code works in most cases. But under some conditions, I cannot get any sound out of the device until I go and toggle the "Matrix 01 Input" and "Matrix 02 Input" enums up and then down (from PCM 1 and PCM 2). This appears to me to be some kind of cache invalidation bug, where the device is out of sync with kernel mixer state.
I will take a look at the code myself at some point, but not for a while. But I am happy to try patches if anyone comes up with anything in the mean time.
Does this change anything (untested)? And are non-enum settings affected?
diff --git a/sound/usb/mixer_scarlett.c b/sound/usb/mixer_scarlett.c index 7438e7c..ad4a452 100644 --- a/sound/usb/mixer_scarlett.c +++ b/sound/usb/mixer_scarlett.c @@ -438,11 +438,9 @@ static int scarlett_ctl_enum_put(struct snd_kcontrol *kctl,
val = ucontrol->value.integer.value[0]; val = val + opt->start;
if (val != oval) {snd_usb_set_cur_mix_value(elem, 0, 0, val);return 1;}return 0;
/* always set cur mix value */snd_usb_set_cur_mix_value(elem, 0, 0, val);return 1;}
static int scarlett_ctl_enum_resume(struct usb_mixer_elem_list *list)
--chris
I had to similarly remove the other instances of (val != oval) in the code, and then it worked as expected.
My guess is that the ALSA mixer settings are cached too soon, before the device itself internally overwrites the settings from NVRAM. We are either missing some invalidation event from the device itself, or we don't do the right thing during initialization.
Adam
Does it happen after some S3/S4 or even during a normal operation without power-saving? The USB-audio device supports autopm, so this needs to be checked, too.
Takashi